Agreed. Most providers appear to simply slap on the "Cloud" sticker to whatever VPS solution they were providing earlier with absolutely no change in features. These are the features that I believe should be present in any provider that claim themselves to be cloud providers:
1. Ability to scale up or down instances with change in demand.
2. Absolutely zero machine agnostic. I really should be bothered whether I am running a linux/windows/lego machine.
3. Auto configuration and notification - Leave little configuration to me so that I can concentrate more on the stuff I like. But, please do notify me in case the whole system burns down.
4. Easily available APIs that are globally used by all - a user account management, cache management, web UI framework etc.
If you see, there are only a few couple of sites (Google App Engine, Heroku et al.) that satisfy these conditions. Rest all are VPS wolfs on Cloud sheep suits.
EC2 is the only provider where you can make API calls to launch 50 servers, have them all booted in a few minutes, then shut them down an hour later. Slicehost's management console will barely load if you have 100 instances, and your servers are all in a single data center. Ditto for Rackspace Cloud. (Both have multiple data centers, but each account is bound to one DC.) I haven't tested Linode beyond booting a single instance on it, but since they bill by the month I assume they're another cloud-ish VPS provider.
Of course, most individuals and small companies would do best to pick a cloud-ish VPS provider. They have much better support than EC2 and small companies usually don't need hundreds of servers.
While Linode's billing is monthly by default, it's also prorated. This means that when you remove a Linode from your account, the account is issued a prorated service credit for the unused time in the billing period. Service credit is always used for further services before charging the card on file.
In other words, you can spin up multiple Linodes, remove them the next day, and all you're really paying for is the day you had them deployed. There's a document on Linode billing located here: http://library.linode.com/linode-platform/billing/
EC2 is the only provider where you can make API calls to launch 50 servers, have them all booted in a few minutes, then shut them down an hour later
GoGrid (http://www.gogrid.com) also has this ability (they also have cloud-based F5 load balancing and mountable block-level storage). They have datacenters on the West and East coasts, and you can pick one when starting an instance.
In my experience Rackspace is good at providing customer support on top of single managed server instances. Their Cloud Sites and Cloud Files offerings (akin to ec2 and s3) feel less than stable, and their management interface encounters unrecoverable errors many times each session. In their Chicago data center they don't allow moving backup snapshots to s3-like storage, all backups are stored with the server. This alone should be a red flag to anyone who is doing any cloud hosting. It just feels like something they're trying to get into, instead of a core competency.
Amazon on the other hand has a much more diverse offering that feels at least somewhat battle tested and polished. They offer you the ability to clone a running server with single click, backups to s3, CDN, block stores, map-reduce, etc. They've also been around for a while, which isn't proof of a good product but it gives you a better feeling that they're serious about the business of cloud hosting (not just hosting).
Not that Amazon doesn't also have its issues. I've heard multiple reports of servers simply "disappearing", which is worrisome.
This disappearing does not happen as much anymore and they now have a lot of different tools which make recovery much faster. After awhile, you find that some of the things you need to do to protect yourself against such failures is good practice in general and using Amazon probably makes your site more reliable overall. In comparison, when I used SoftLayer, a hardware failure took down a site I had for 4+ hours while an engineer went to the machine, found out what was wrong, fixed it, etc. On Amazon, the same thing would take minutes to recover from.
Having just gone through evaluating various cloud providers for our high-traffic website, this is my personal definition: utility computing as electricity, meaning unlimited (in both directions) metered access. Most "cloud" companies offer neither: a typical VPS provider requires a contract that stipulates exactly how much you may consume, and when your consumption changes, you pay and wait. Imagine if you had to do that at home... a $25 set up fee and 48 hours to turn on an extra light.
Rackspace Cloud Sites would be a great example of cloud computing. It behaves like traditional shared hosting, with an easy to use control panel, but behind the scenes it can scale automatically with traffic and load. Obviously the database doesn't scale, but the frontend web servers do, all without user intervention.
EDIT: to the downmodders, please explain how Cloud Sites is not an example of cloud computing. It is a pool of shared hardware and software resources, the details and maintenance of which are abastracted away from the users, with the ability to automatically and infinitely scale up based on demand without user intervention, sold using an IaaS model?
Rackspace, VPS.NET, Linode et al are all just VPS++marketing (_not_ cloud).